COMMENT ON THE THEORETICAL AND EMPIRICAL EVIDENCE OF FUNDAMENTAL INDEXING

Jeffrey Graham

In a recent article in this journal, Arnott and Hsu [Journal of Investment Management 6(1), 2008] proposed an equity pricing model in which a security’s current market price can help predict its future return. Their basic framework assumed that current prices consist of an underlying “fair value” and a mean-reverting “pricing error” that are not observable to investors. Their main finding was that securities with a higher price will tend to have below normal returns, thus a cap-weighted index that places more weight on high-priced securities will also tend to have below normal returns. This note describes how this result hinges on one very strong implicit assumption. In order to derive this result, Arnott and Hsu have assumed that investors have some reference point in which they know the stock’s fair value. This directly contradicts the notion that fair values are unobservable. I show that if one relaxes this assumption the result does not necessarily follow. In addition, I provide some empirical evidence suggesting that on average cap-weighted indexes have not experienced lower returns due to the systematic over-pricing of large cap securities, as Arnott and Hsu predict.